Coordination-Free Replicated Datalog Streams with Application-Specific Availability

Abstract

Data are continuously generated, processed and consumed everywhere. However, the computing devices may be disconnected from the network. We want the disconnection to have as little impact on the availability of data as possible. This paper presents an approach to transform a sequential Datalog program into a distributed one. The distributed program meets the specified availability and fault-tolerance requirements. The continuous execution of the distributed program is coordination free and eventual consistent. To achieve this, we augment the Datalog program with a composite CRDT (Conflict-free Replicated Data Type). We have implemented a prototype and run experiments to study the performance characteristics of the approach.